Firefox 122.0 released Posted Jan 23, 2024 17:40 UTC (Tue) by cyperpunks (subscriber, #39406) [Link] (32 responses) > and a shiny new .deb package. This is in fact important these days, as Ubuntu has stopped doing their main purpose in life: creating .deb packages for popular software. Firefox 122.0 released Posted Jan 23, 2024 17:46 UTC (Tue) by dskoll (subscriber, #1630) [Link] (23 responses) Neat! I'm on Debian, but usually just install the Linux binary tarball. I might evaluate switching to the .deb. Firefox 122.0 released Posted Jan 23, 2024 20:01 UTC (Tue) by jfebrer (guest, #82539) [Link] (22 responses) Here are some reasons: https://blog.mozilla.org/en/products/4-reasons-to-try-moz... Firefox 122.0 released Posted Jan 23, 2024 20:29 UTC (Tue) by ballombe (subscriber, #9523) [Link] (21 responses) 3. Better performance: this one is shameful: - no hard number. For all I know, distro packages might actually be faster. - gratuitously casting aspersion on packages maintainers. - ridiculous statement about open-source. - making assertions which are difficult to reconcile with the fact they provide a single binary for all distros. Firefox 122.0 released Posted Jan 23, 2024 20:38 UTC (Tue) by dskoll (subscriber, #1630) [Link] (19 responses) Yes, the performance claim seems somewhat dubious. Also, I download their distro-agnostic tarball and presumably that's built with the same compiler optimizations. I've scripted something that checks for new releases and updates with just one command, so I guess I'll keep doing that. Firefox 122.0 released Posted Jan 23, 2024 21:27 UTC (Tue) by pizza (subscriber, #46) [Link] (18 responses) > Yes, the performance claim seems somewhat dubious. Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) This is also why Mozilla's windows binaries, when run under Wine, noticably outperform native distro-provided binaries. (Note distros also typically perform builds with various hardening strategies enabled, which Mozilla's own builds presumably don't do. This would adversely affect performance, though many would consider this to be an acceptable trade-off) Firefox 122.0 released Posted Jan 23, 2024 21:42 UTC (Tue) by ballombe (subscriber, #9523) [Link] (15 responses) > Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) There is nothing that prevent distro to use multi-pass profile-guided optimization... Firefox 122.0 released Posted Jan 23, 2024 21:48 UTC (Tue) by mb (subscriber, #50428) [Link] > There is nothing that prevent distro to use multi-pass profile-guided optimization... Except for build hardware constraints. Build time is not for free. Firefox 122.0 released Posted Jan 23, 2024 22:16 UTC (Tue) by pizza (subscriber, #46) [Link] (12 responses) > There is nothing that prevent distro to use multi-pass profile-guided optimization... At minimum, it doubles the build time. But it also takes additional time/resources to run the benchmark/test/etc suites used to generate the profile data to feed back into the second pass. Plus there's the need to create/maintain the necessary environment to _run_ those suites -- presumably requiring connecting to server(s) containing representative content. This poses additional issues as some (many? most?) distros disable network access during builds. Sure, all of this is _doable_ but it's, at minimum, a considerable amount of work to set up and maintain. Plus the doubling (or so) of build times. (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) Firefox 122.0 released Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link] > (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) *chromium's 5x as long build time gives a nice friendly wave* Firefox 122.0 released Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses) In Gentoo, there's a "pho" flag to do this. Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though. Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol Firefox 122.0 released Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses) This is a strange bet coming from a m68k porter! firefox Build-Space: 38435228 Build-Time: 4843 chromium Build-Space: 13361912 Build-Time: 31142 libreoffice Build-Space: 47828428 Build-Time: 20769 llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953 Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build. Firefox 122.0 released Posted Jan 24, 2024 6:02 UTC (Wed) by roc (subscriber, #30627) [Link] Maybe not, but the important question is: do they *actually do* them? Firefox 122.0 released Posted Jan 24, 2024 9:05 UTC (Wed) by RX14 (subscriber, #123970) [Link] (1 responses) Arch Linux has been doing PGO builds of firefox ever since i first read the PKGBUILD. To hear that debian/ubuntu wouldn't, when the process is automated and they are larger distros, is strange. Firefox 122.0 released Posted Jan 24, 2024 9:44 UTC (Wed) by cyperpunks (subscriber, #39406) [Link] Fedora also do PGO build of Firefox. It's better that one build machine do double work than millions of machines doing 10% extra work. Firefox 122.0 released Posted Jan 23, 2024 23:22 UTC (Tue) by jmalcolm (subscriber, #8876) [Link] Pretty sure this is a shot at the Snap package. I do not use Ubuntu but I have read many reports that the Snap performs quite badly. That is probably one of the primary motivations for releasing this .deb for Ubuntu. As for Debian, depending on which version you choose, the official package may be quite a bit out of date. For Debian, that would be the main motivation here. The .deb is only for Ubuntu and Debian systems. Both of the above ring true to me and fit nicely into what Mozilla is saying. In the official release notes though, Mozilla actually recommends sticking with your distro package: https://support.mozilla.org/en-US/kb/install-firefox-linu... Nothing "shameful" here that I can see. Firefox 122.0 released Posted Jan 23, 2024 18:17 UTC (Tue) by mb (subscriber, #50428) [Link] (4 responses) Yeah, that's nice. Kind of. I'm unsure, though. I currently use the flatpak and it mostly works fine. It has a few OS integration issues (e.g. fonts), but it's minor enough that I can live with it. Now, the deb will probably integrate better into the OS. At least I hope so. But what I really dislike is the proposal to prioritize the mozilla repo: >Configure APT to prioritize packages from the Mozilla repository: >Package: * >Pin: origin packages.mozilla.org >Pin-Priority: 1000 That basically means mozilla can install and override any package on my system, right? I mean, I'm not so much worried about hostile packages. If Firefox is hostile, that's basically game over already. I'm more worried about mistakes. Or that they start updating random libraries because they need it, causing conflicts with base install packages. What do you think? Is it worth going from flatpak to deb? Or is flatpak better? Firefox 122.0 released Posted Jan 23, 2024 19:08 UTC (Tue) by Luban (guest, #98071) [Link] You should be able to pin only specific package(s), instead of '*', to the specific origin: https://wiki.debian.org/AptConfiguration#Prevent.2Fselect... Firefox 122.0 released Posted Jan 24, 2024 16:01 UTC (Wed) by lproven (guest, #110432) [Link] (2 responses) > But what I really dislike is the proposal to prioritize the mozilla repo This is necessary on Ubuntu, and I think that Ubuntu is the primary goal here. To explain: It's necessary because Ubuntu has a dummy Firefox .deb package which it uses to automatically prefer the Snap-packaged version. So, if you remove the snap-packaged Ubuntu version of Firefox, then add an alternative repository, then do `apt update ; apt install firefox` it will re-install the snap package. This confuses non-experts, and non-experts are the #1 target audience of Ubuntu. To bypass this, you have to set the new repo's priority as higher than the distro's own. I think it's aimed at Ubuntu because: [1] Debian includes Firefox anyway (albeit the ESR version), packaged as a .deb. [2] Linux Mint, probably the #1 Ubuntu downstream, already includes a Mozilla-codeveloped native .deb-packaged Firefox. [3] Of other Ubuntu downstreams, Linux Lite uses Chrome by default, and I *think* Zorin uses the Firefox Flatpak, as it ships several tens of gigabytes of flatpak-packaged apps as standard. (Yes, really that big: Zorin OS is the single biggest distro I've ever seen. Some other distros require as much or more disk, e.g. Deepin, because of redundant failover root partitions; Zorin OS just needs it for all the flatpak apps.) Firefox 122.0 released Posted Jan 25, 2024 4:34 UTC (Thu) by interalia (subscriber, #26615) [Link] (1 responses) > [1] Debian includes Firefox anyway (albeit the ESR version), packaged as a .deb. Eh? There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable. Firefox 122.0 released Posted Jan 25, 2024 7:23 UTC (Thu) by zaitseff (subscriber, #851) [Link] > There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable. Debian Sid has both packages (firefox and firefox-esr), but only firefox-esr can progress to Debian Testing and from there to an actual release. Bug #817954 ensures that this is the case: https://bugs.debian.org/817954 Firefox 122.0 released Posted Jan 23, 2024 19:19 UTC (Tue) by zaitseff (subscriber, #851) [Link] (1 responses) > [...] as Ubuntu has stopped doing their main purpose in life: creating .deb packages for popular software. Which is why I recompile Debian's firefox, firefox-esr and chromium packages for Ubuntu at https://ftp.zap.org.au/pub/debian/ (see the README file at that location for details), with patches and backports as needed to make this happen. I bump the epoch of the package to make it override Ubuntu's versions. What interests me are answers to the following questions: 1. Does Mozilla provide the equivalent deb-src line for package sources? Too many upstream providers of DEB packages do NOT have such sources -- which means recreating their DEB packages is either non-trivial or impossible without access to their build infrastructure. 2. What patches from the current Debian package, if any, are included? These patches (which you can see at https://salsa.debian.org/mozilla-team/firefox/-/tree/rele...) are mostly build-time fixes, but some might be useful upstream (eg, Set-DPI-to-system-settings.patch). Firefox 122.0 released Posted Jan 23, 2024 20:39 UTC (Tue) by ballombe (subscriber, #9523) [Link] > 1. Does Mozilla provide the equivalent deb-src line for package sources? Too many upstream providers of DEB packages do NOT have such sources -- which means recreating their DEB packages is either non-trivial or impossible without access to their build infrastructure. No they do not, as far as I can see. Firefox 122.0 released Posted Jan 23, 2024 19:19 UTC (Tue) by ballombe (subscriber, #9523) [Link] The repository is do not allow directory listing which makes it harder to see what they actually provide. So far it seems they only provide amd64 and i386. But more importantly: they ship a single build for all debian/ubuntu releases (the packages depends on libc >= 2.17) so I would not expect any kind of system integration. Firefox 122.0 released Posted Jan 23, 2024 21:33 UTC (Tue) by atai (subscriber, #10977) [Link] (1 responses) Does anyone know for sure instructions to disable snapd so Firefox .deb's can take priority when installation or upgrade, on Ubuntu? I disabled and removed snap package but when upgrading software the Firefox snap package comes back. Very annoying Firefox 122.0 released Posted Jan 24, 2024 2:19 UTC (Wed) by PengZheng (subscriber, #108006) [Link] The 5th step of the official guide should fix the priority issue: https://support.mozilla.org/en-US/kb/install-firefox-linu... echo ' Package: * Pin: origin packages.mozilla.org Pin-Priority: 1000 ' | sudo tee /etc/apt/preferences.d/mozilla Firefox 122.0 released Posted Jan 24, 2024 5:11 UTC (Wed) by Cyberax (✭ supporter ✭, #52523) [Link] Looks like passkeys finally work with BitWarden! Super nice! Firefox 122.0 released Posted Jan 24, 2024 7:17 UTC (Wed) by donbarry (guest, #10485) [Link] (2 responses) But does the deb allow installing xpi files without official signing by Mozilla? Some of us refuse to play the walled garden game and give an external provider the right to determine how we configure our software or even install it. If not, then their increasingly hypocritical invocations of "open source" simply expose the gulf between the claims and the practice. Firefox 122.0 released Posted Jan 26, 2024 1:23 UTC (Fri) by higuita (guest, #32245) [Link] (1 responses) The digital sign is there to protect everyone, that avoids normal people from installing trojan extensions. Now we have few attacks using it exactly because of the required signature, but in the past, there were many fake extensions for firefox... with most extensions being opensource, someone can pull the source and adding some extra code and name it the same or similar and will always catch someone. There are ways to install unsigned packages in official firefox, but those require several extra steps, so not something that people would do by accident. Or use the developer versions or nightly and change the xpinstall.signatures.required setting, those can allow unsigned extensions, usually for developer testing, but if you want to take the risk, use those versions Firefox 122.0 released Posted Jan 29, 2024 15:10 UTC (Mon) by intelfx (subscriber, #130118) [Link] > There are ways to install unsigned packages in official firefox, but those require several extra steps, so not something that people would do by accident Are there? I'm genuinely interested. Firefox 122.0 released Posted Jan 24, 2024 17:26 UTC (Wed) by traxus (guest, #119726) [Link] I am on Ubuntu 23.10 and could tell a speed difference between the .deb package provided by Mozilla and the Launchpad PPA I use to install Firefox. All around, things are noticeably faster *for me* from startup, tabs, and using Bitwarden.
Posted Jan 23, 2024 17:40 UTC (Tue) by cyperpunks (subscriber, #39406) [Link] (32 responses)
This is in fact important these days, as Ubuntu has stopped doing their main purpose in life: creating .deb packages for popular software.
Firefox 122.0 released Posted Jan 23, 2024 17:46 UTC (Tue) by dskoll (subscriber, #1630) [Link] (23 responses) Neat! I'm on Debian, but usually just install the Linux binary tarball. I might evaluate switching to the .deb. Firefox 122.0 released Posted Jan 23, 2024 20:01 UTC (Tue) by jfebrer (guest, #82539) [Link] (22 responses) Here are some reasons: https://blog.mozilla.org/en/products/4-reasons-to-try-moz... Firefox 122.0 released Posted Jan 23, 2024 20:29 UTC (Tue) by ballombe (subscriber, #9523) [Link] (21 responses) 3. Better performance: this one is shameful: - no hard number. For all I know, distro packages might actually be faster. - gratuitously casting aspersion on packages maintainers. - ridiculous statement about open-source. - making assertions which are difficult to reconcile with the fact they provide a single binary for all distros. Firefox 122.0 released Posted Jan 23, 2024 20:38 UTC (Tue) by dskoll (subscriber, #1630) [Link] (19 responses) Yes, the performance claim seems somewhat dubious. Also, I download their distro-agnostic tarball and presumably that's built with the same compiler optimizations. I've scripted something that checks for new releases and updates with just one command, so I guess I'll keep doing that. Firefox 122.0 released Posted Jan 23, 2024 21:27 UTC (Tue) by pizza (subscriber, #46) [Link] (18 responses) > Yes, the performance claim seems somewhat dubious. Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) This is also why Mozilla's windows binaries, when run under Wine, noticably outperform native distro-provided binaries. (Note distros also typically perform builds with various hardening strategies enabled, which Mozilla's own builds presumably don't do. This would adversely affect performance, though many would consider this to be an acceptable trade-off) Firefox 122.0 released Posted Jan 23, 2024 21:42 UTC (Tue) by ballombe (subscriber, #9523) [Link] (15 responses) > Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) There is nothing that prevent distro to use multi-pass profile-guided optimization... Firefox 122.0 released Posted Jan 23, 2024 21:48 UTC (Tue) by mb (subscriber, #50428) [Link] > There is nothing that prevent distro to use multi-pass profile-guided optimization... Except for build hardware constraints. Build time is not for free. Firefox 122.0 released Posted Jan 23, 2024 22:16 UTC (Tue) by pizza (subscriber, #46) [Link] (12 responses) > There is nothing that prevent distro to use multi-pass profile-guided optimization... At minimum, it doubles the build time. But it also takes additional time/resources to run the benchmark/test/etc suites used to generate the profile data to feed back into the second pass. Plus there's the need to create/maintain the necessary environment to _run_ those suites -- presumably requiring connecting to server(s) containing representative content. This poses additional issues as some (many? most?) distros disable network access during builds. Sure, all of this is _doable_ but it's, at minimum, a considerable amount of work to set up and maintain. Plus the doubling (or so) of build times. (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) Firefox 122.0 released Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link] > (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) *chromium's 5x as long build time gives a nice friendly wave* Firefox 122.0 released Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses) In Gentoo, there's a "pho" flag to do this. Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though. Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol Firefox 122.0 released Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses) This is a strange bet coming from a m68k porter! firefox Build-Space: 38435228 Build-Time: 4843 chromium Build-Space: 13361912 Build-Time: 31142 libreoffice Build-Space: 47828428 Build-Time: 20769 llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953 Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build. Firefox 122.0 released Posted Jan 24, 2024 6:02 UTC (Wed) by roc (subscriber, #30627) [Link] Maybe not, but the important question is: do they *actually do* them? Firefox 122.0 released Posted Jan 24, 2024 9:05 UTC (Wed) by RX14 (subscriber, #123970) [Link] (1 responses) Arch Linux has been doing PGO builds of firefox ever since i first read the PKGBUILD. To hear that debian/ubuntu wouldn't, when the process is automated and they are larger distros, is strange. Firefox 122.0 released Posted Jan 24, 2024 9:44 UTC (Wed) by cyperpunks (subscriber, #39406) [Link] Fedora also do PGO build of Firefox. It's better that one build machine do double work than millions of machines doing 10% extra work. Firefox 122.0 released Posted Jan 23, 2024 23:22 UTC (Tue) by jmalcolm (subscriber, #8876) [Link] Pretty sure this is a shot at the Snap package. I do not use Ubuntu but I have read many reports that the Snap performs quite badly. That is probably one of the primary motivations for releasing this .deb for Ubuntu. As for Debian, depending on which version you choose, the official package may be quite a bit out of date. For Debian, that would be the main motivation here. The .deb is only for Ubuntu and Debian systems. Both of the above ring true to me and fit nicely into what Mozilla is saying. In the official release notes though, Mozilla actually recommends sticking with your distro package: https://support.mozilla.org/en-US/kb/install-firefox-linu... Nothing "shameful" here that I can see. Firefox 122.0 released Posted Jan 23, 2024 18:17 UTC (Tue) by mb (subscriber, #50428) [Link] (4 responses) Yeah, that's nice. Kind of. I'm unsure, though. I currently use the flatpak and it mostly works fine. It has a few OS integration issues (e.g. fonts), but it's minor enough that I can live with it. Now, the deb will probably integrate better into the OS. At least I hope so. But what I really dislike is the proposal to prioritize the mozilla repo: >Configure APT to prioritize packages from the Mozilla repository: >Package: * >Pin: origin packages.mozilla.org >Pin-Priority: 1000 That basically means mozilla can install and override any package on my system, right? I mean, I'm not so much worried about hostile packages. If Firefox is hostile, that's basically game over already. I'm more worried about mistakes. Or that they start updating random libraries because they need it, causing conflicts with base install packages. What do you think? Is it worth going from flatpak to deb? Or is flatpak better? Firefox 122.0 released Posted Jan 23, 2024 19:08 UTC (Tue) by Luban (guest, #98071) [Link] You should be able to pin only specific package(s), instead of '*', to the specific origin: https://wiki.debian.org/AptConfiguration#Prevent.2Fselect... Firefox 122.0 released Posted Jan 24, 2024 16:01 UTC (Wed) by lproven (guest, #110432) [Link] (2 responses) > But what I really dislike is the proposal to prioritize the mozilla repo This is necessary on Ubuntu, and I think that Ubuntu is the primary goal here. To explain: It's necessary because Ubuntu has a dummy Firefox .deb package which it uses to automatically prefer the Snap-packaged version. So, if you remove the snap-packaged Ubuntu version of Firefox, then add an alternative repository, then do `apt update ; apt install firefox` it will re-install the snap package. This confuses non-experts, and non-experts are the #1 target audience of Ubuntu. To bypass this, you have to set the new repo's priority as higher than the distro's own. I think it's aimed at Ubuntu because: [1] Debian includes Firefox anyway (albeit the ESR version), packaged as a .deb. [2] Linux Mint, probably the #1 Ubuntu downstream, already includes a Mozilla-codeveloped native .deb-packaged Firefox. [3] Of other Ubuntu downstreams, Linux Lite uses Chrome by default, and I *think* Zorin uses the Firefox Flatpak, as it ships several tens of gigabytes of flatpak-packaged apps as standard. (Yes, really that big: Zorin OS is the single biggest distro I've ever seen. Some other distros require as much or more disk, e.g. Deepin, because of redundant failover root partitions; Zorin OS just needs it for all the flatpak apps.) Firefox 122.0 released Posted Jan 25, 2024 4:34 UTC (Thu) by interalia (subscriber, #26615) [Link] (1 responses) > [1] Debian includes Firefox anyway (albeit the ESR version), packaged as a .deb. Eh? There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable. Firefox 122.0 released Posted Jan 25, 2024 7:23 UTC (Thu) by zaitseff (subscriber, #851) [Link] > There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable. Debian Sid has both packages (firefox and firefox-esr), but only firefox-esr can progress to Debian Testing and from there to an actual release. Bug #817954 ensures that this is the case: https://bugs.debian.org/817954 Firefox 122.0 released Posted Jan 23, 2024 19:19 UTC (Tue) by zaitseff (subscriber, #851) [Link] (1 responses) > [...] as Ubuntu has stopped doing their main purpose in life: creating .deb packages for popular software. Which is why I recompile Debian's firefox, firefox-esr and chromium packages for Ubuntu at https://ftp.zap.org.au/pub/debian/ (see the README file at that location for details), with patches and backports as needed to make this happen. I bump the epoch of the package to make it override Ubuntu's versions. What interests me are answers to the following questions: 1. Does Mozilla provide the equivalent deb-src line for package sources? Too many upstream providers of DEB packages do NOT have such sources -- which means recreating their DEB packages is either non-trivial or impossible without access to their build infrastructure. 2. What patches from the current Debian package, if any, are included? These patches (which you can see at https://salsa.debian.org/mozilla-team/firefox/-/tree/rele...) are mostly build-time fixes, but some might be useful upstream (eg, Set-DPI-to-system-settings.patch). Firefox 122.0 released Posted Jan 23, 2024 20:39 UTC (Tue) by ballombe (subscriber, #9523) [Link] > 1. Does Mozilla provide the equivalent deb-src line for package sources? Too many upstream providers of DEB packages do NOT have such sources -- which means recreating their DEB packages is either non-trivial or impossible without access to their build infrastructure. No they do not, as far as I can see. Firefox 122.0 released Posted Jan 23, 2024 19:19 UTC (Tue) by ballombe (subscriber, #9523) [Link] The repository is do not allow directory listing which makes it harder to see what they actually provide. So far it seems they only provide amd64 and i386. But more importantly: they ship a single build for all debian/ubuntu releases (the packages depends on libc >= 2.17) so I would not expect any kind of system integration.
Posted Jan 23, 2024 17:46 UTC (Tue) by dskoll (subscriber, #1630) [Link] (23 responses)
Neat! I'm on Debian, but usually just install the Linux binary tarball. I might evaluate switching to the .deb.
Firefox 122.0 released Posted Jan 23, 2024 20:01 UTC (Tue) by jfebrer (guest, #82539) [Link] (22 responses) Here are some reasons: https://blog.mozilla.org/en/products/4-reasons-to-try-moz... Firefox 122.0 released Posted Jan 23, 2024 20:29 UTC (Tue) by ballombe (subscriber, #9523) [Link] (21 responses) 3. Better performance: this one is shameful: - no hard number. For all I know, distro packages might actually be faster. - gratuitously casting aspersion on packages maintainers. - ridiculous statement about open-source. - making assertions which are difficult to reconcile with the fact they provide a single binary for all distros. Firefox 122.0 released Posted Jan 23, 2024 20:38 UTC (Tue) by dskoll (subscriber, #1630) [Link] (19 responses) Yes, the performance claim seems somewhat dubious. Also, I download their distro-agnostic tarball and presumably that's built with the same compiler optimizations. I've scripted something that checks for new releases and updates with just one command, so I guess I'll keep doing that. Firefox 122.0 released Posted Jan 23, 2024 21:27 UTC (Tue) by pizza (subscriber, #46) [Link] (18 responses) > Yes, the performance claim seems somewhat dubious. Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) This is also why Mozilla's windows binaries, when run under Wine, noticably outperform native distro-provided binaries. (Note distros also typically perform builds with various hardening strategies enabled, which Mozilla's own builds presumably don't do. This would adversely affect performance, though many would consider this to be an acceptable trade-off) Firefox 122.0 released Posted Jan 23, 2024 21:42 UTC (Tue) by ballombe (subscriber, #9523) [Link] (15 responses) > Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) There is nothing that prevent distro to use multi-pass profile-guided optimization... Firefox 122.0 released Posted Jan 23, 2024 21:48 UTC (Tue) by mb (subscriber, #50428) [Link] > There is nothing that prevent distro to use multi-pass profile-guided optimization... Except for build hardware constraints. Build time is not for free. Firefox 122.0 released Posted Jan 23, 2024 22:16 UTC (Tue) by pizza (subscriber, #46) [Link] (12 responses) > There is nothing that prevent distro to use multi-pass profile-guided optimization... At minimum, it doubles the build time. But it also takes additional time/resources to run the benchmark/test/etc suites used to generate the profile data to feed back into the second pass. Plus there's the need to create/maintain the necessary environment to _run_ those suites -- presumably requiring connecting to server(s) containing representative content. This poses additional issues as some (many? most?) distros disable network access during builds. Sure, all of this is _doable_ but it's, at minimum, a considerable amount of work to set up and maintain. Plus the doubling (or so) of build times. (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) Firefox 122.0 released Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link] > (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) *chromium's 5x as long build time gives a nice friendly wave* Firefox 122.0 released Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses) In Gentoo, there's a "pho" flag to do this. Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though. Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol Firefox 122.0 released Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses) This is a strange bet coming from a m68k porter! firefox Build-Space: 38435228 Build-Time: 4843 chromium Build-Space: 13361912 Build-Time: 31142 libreoffice Build-Space: 47828428 Build-Time: 20769 llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953 Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build. Firefox 122.0 released Posted Jan 24, 2024 6:02 UTC (Wed) by roc (subscriber, #30627) [Link] Maybe not, but the important question is: do they *actually do* them? Firefox 122.0 released Posted Jan 24, 2024 9:05 UTC (Wed) by RX14 (subscriber, #123970) [Link] (1 responses) Arch Linux has been doing PGO builds of firefox ever since i first read the PKGBUILD. To hear that debian/ubuntu wouldn't, when the process is automated and they are larger distros, is strange. Firefox 122.0 released Posted Jan 24, 2024 9:44 UTC (Wed) by cyperpunks (subscriber, #39406) [Link] Fedora also do PGO build of Firefox. It's better that one build machine do double work than millions of machines doing 10% extra work. Firefox 122.0 released Posted Jan 23, 2024 23:22 UTC (Tue) by jmalcolm (subscriber, #8876) [Link] Pretty sure this is a shot at the Snap package. I do not use Ubuntu but I have read many reports that the Snap performs quite badly. That is probably one of the primary motivations for releasing this .deb for Ubuntu. As for Debian, depending on which version you choose, the official package may be quite a bit out of date. For Debian, that would be the main motivation here. The .deb is only for Ubuntu and Debian systems. Both of the above ring true to me and fit nicely into what Mozilla is saying. In the official release notes though, Mozilla actually recommends sticking with your distro package: https://support.mozilla.org/en-US/kb/install-firefox-linu... Nothing "shameful" here that I can see.
Posted Jan 23, 2024 20:01 UTC (Tue) by jfebrer (guest, #82539) [Link] (22 responses)
https://blog.mozilla.org/en/products/4-reasons-to-try-moz...
Firefox 122.0 released Posted Jan 23, 2024 20:29 UTC (Tue) by ballombe (subscriber, #9523) [Link] (21 responses) 3. Better performance: this one is shameful: - no hard number. For all I know, distro packages might actually be faster. - gratuitously casting aspersion on packages maintainers. - ridiculous statement about open-source. - making assertions which are difficult to reconcile with the fact they provide a single binary for all distros. Firefox 122.0 released Posted Jan 23, 2024 20:38 UTC (Tue) by dskoll (subscriber, #1630) [Link] (19 responses) Yes, the performance claim seems somewhat dubious. Also, I download their distro-agnostic tarball and presumably that's built with the same compiler optimizations. I've scripted something that checks for new releases and updates with just one command, so I guess I'll keep doing that. Firefox 122.0 released Posted Jan 23, 2024 21:27 UTC (Tue) by pizza (subscriber, #46) [Link] (18 responses) > Yes, the performance claim seems somewhat dubious. Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) This is also why Mozilla's windows binaries, when run under Wine, noticably outperform native distro-provided binaries. (Note distros also typically perform builds with various hardening strategies enabled, which Mozilla's own builds presumably don't do. This would adversely affect performance, though many would consider this to be an acceptable trade-off) Firefox 122.0 released Posted Jan 23, 2024 21:42 UTC (Tue) by ballombe (subscriber, #9523) [Link] (15 responses) > Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) There is nothing that prevent distro to use multi-pass profile-guided optimization... Firefox 122.0 released Posted Jan 23, 2024 21:48 UTC (Tue) by mb (subscriber, #50428) [Link] > There is nothing that prevent distro to use multi-pass profile-guided optimization... Except for build hardware constraints. Build time is not for free. Firefox 122.0 released Posted Jan 23, 2024 22:16 UTC (Tue) by pizza (subscriber, #46) [Link] (12 responses) > There is nothing that prevent distro to use multi-pass profile-guided optimization... At minimum, it doubles the build time. But it also takes additional time/resources to run the benchmark/test/etc suites used to generate the profile data to feed back into the second pass. Plus there's the need to create/maintain the necessary environment to _run_ those suites -- presumably requiring connecting to server(s) containing representative content. This poses additional issues as some (many? most?) distros disable network access during builds. Sure, all of this is _doable_ but it's, at minimum, a considerable amount of work to set up and maintain. Plus the doubling (or so) of build times. (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) Firefox 122.0 released Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link] > (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) *chromium's 5x as long build time gives a nice friendly wave* Firefox 122.0 released Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses) In Gentoo, there's a "pho" flag to do this. Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though. Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol Firefox 122.0 released Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses) This is a strange bet coming from a m68k porter! firefox Build-Space: 38435228 Build-Time: 4843 chromium Build-Space: 13361912 Build-Time: 31142 libreoffice Build-Space: 47828428 Build-Time: 20769 llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953 Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build. Firefox 122.0 released Posted Jan 24, 2024 6:02 UTC (Wed) by roc (subscriber, #30627) [Link] Maybe not, but the important question is: do they *actually do* them? Firefox 122.0 released Posted Jan 24, 2024 9:05 UTC (Wed) by RX14 (subscriber, #123970) [Link] (1 responses) Arch Linux has been doing PGO builds of firefox ever since i first read the PKGBUILD. To hear that debian/ubuntu wouldn't, when the process is automated and they are larger distros, is strange. Firefox 122.0 released Posted Jan 24, 2024 9:44 UTC (Wed) by cyperpunks (subscriber, #39406) [Link] Fedora also do PGO build of Firefox. It's better that one build machine do double work than millions of machines doing 10% extra work. Firefox 122.0 released Posted Jan 23, 2024 23:22 UTC (Tue) by jmalcolm (subscriber, #8876) [Link] Pretty sure this is a shot at the Snap package. I do not use Ubuntu but I have read many reports that the Snap performs quite badly. That is probably one of the primary motivations for releasing this .deb for Ubuntu. As for Debian, depending on which version you choose, the official package may be quite a bit out of date. For Debian, that would be the main motivation here. The .deb is only for Ubuntu and Debian systems. Both of the above ring true to me and fit nicely into what Mozilla is saying. In the official release notes though, Mozilla actually recommends sticking with your distro package: https://support.mozilla.org/en-US/kb/install-firefox-linu... Nothing "shameful" here that I can see.
Posted Jan 23, 2024 20:29 UTC (Tue) by ballombe (subscriber, #9523) [Link] (21 responses)
Firefox 122.0 released Posted Jan 23, 2024 20:38 UTC (Tue) by dskoll (subscriber, #1630) [Link] (19 responses) Yes, the performance claim seems somewhat dubious. Also, I download their distro-agnostic tarball and presumably that's built with the same compiler optimizations. I've scripted something that checks for new releases and updates with just one command, so I guess I'll keep doing that. Firefox 122.0 released Posted Jan 23, 2024 21:27 UTC (Tue) by pizza (subscriber, #46) [Link] (18 responses) > Yes, the performance claim seems somewhat dubious. Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) This is also why Mozilla's windows binaries, when run under Wine, noticably outperform native distro-provided binaries. (Note distros also typically perform builds with various hardening strategies enabled, which Mozilla's own builds presumably don't do. This would adversely affect performance, though many would consider this to be an acceptable trade-off) Firefox 122.0 released Posted Jan 23, 2024 21:42 UTC (Tue) by ballombe (subscriber, #9523) [Link] (15 responses) > Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) There is nothing that prevent distro to use multi-pass profile-guided optimization... Firefox 122.0 released Posted Jan 23, 2024 21:48 UTC (Tue) by mb (subscriber, #50428) [Link] > There is nothing that prevent distro to use multi-pass profile-guided optimization... Except for build hardware constraints. Build time is not for free. Firefox 122.0 released Posted Jan 23, 2024 22:16 UTC (Tue) by pizza (subscriber, #46) [Link] (12 responses) > There is nothing that prevent distro to use multi-pass profile-guided optimization... At minimum, it doubles the build time. But it also takes additional time/resources to run the benchmark/test/etc suites used to generate the profile data to feed back into the second pass. Plus there's the need to create/maintain the necessary environment to _run_ those suites -- presumably requiring connecting to server(s) containing representative content. This poses additional issues as some (many? most?) distros disable network access during builds. Sure, all of this is _doable_ but it's, at minimum, a considerable amount of work to set up and maintain. Plus the doubling (or so) of build times. (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) Firefox 122.0 released Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link] > (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) *chromium's 5x as long build time gives a nice friendly wave* Firefox 122.0 released Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses) In Gentoo, there's a "pho" flag to do this. Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though. Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol Firefox 122.0 released Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses) This is a strange bet coming from a m68k porter! firefox Build-Space: 38435228 Build-Time: 4843 chromium Build-Space: 13361912 Build-Time: 31142 libreoffice Build-Space: 47828428 Build-Time: 20769 llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953 Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build. Firefox 122.0 released Posted Jan 24, 2024 6:02 UTC (Wed) by roc (subscriber, #30627) [Link] Maybe not, but the important question is: do they *actually do* them? Firefox 122.0 released Posted Jan 24, 2024 9:05 UTC (Wed) by RX14 (subscriber, #123970) [Link] (1 responses) Arch Linux has been doing PGO builds of firefox ever since i first read the PKGBUILD. To hear that debian/ubuntu wouldn't, when the process is automated and they are larger distros, is strange. Firefox 122.0 released Posted Jan 24, 2024 9:44 UTC (Wed) by cyperpunks (subscriber, #39406) [Link] Fedora also do PGO build of Firefox. It's better that one build machine do double work than millions of machines doing 10% extra work. Firefox 122.0 released Posted Jan 23, 2024 23:22 UTC (Tue) by jmalcolm (subscriber, #8876) [Link] Pretty sure this is a shot at the Snap package. I do not use Ubuntu but I have read many reports that the Snap performs quite badly. That is probably one of the primary motivations for releasing this .deb for Ubuntu. As for Debian, depending on which version you choose, the official package may be quite a bit out of date. For Debian, that would be the main motivation here. The .deb is only for Ubuntu and Debian systems. Both of the above ring true to me and fit nicely into what Mozilla is saying. In the official release notes though, Mozilla actually recommends sticking with your distro package: https://support.mozilla.org/en-US/kb/install-firefox-linu... Nothing "shameful" here that I can see.
Posted Jan 23, 2024 20:38 UTC (Tue) by dskoll (subscriber, #1630) [Link] (19 responses)
Yes, the performance claim seems somewhat dubious. Also, I download their distro-agnostic tarball and presumably that's built with the same compiler optimizations. I've scripted something that checks for new releases and updates with just one command, so I guess I'll keep doing that.
Firefox 122.0 released Posted Jan 23, 2024 21:27 UTC (Tue) by pizza (subscriber, #46) [Link] (18 responses) > Yes, the performance claim seems somewhat dubious. Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) This is also why Mozilla's windows binaries, when run under Wine, noticably outperform native distro-provided binaries. (Note distros also typically perform builds with various hardening strategies enabled, which Mozilla's own builds presumably don't do. This would adversely affect performance, though many would consider this to be an acceptable trade-off) Firefox 122.0 released Posted Jan 23, 2024 21:42 UTC (Tue) by ballombe (subscriber, #9523) [Link] (15 responses) > Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) There is nothing that prevent distro to use multi-pass profile-guided optimization... Firefox 122.0 released Posted Jan 23, 2024 21:48 UTC (Tue) by mb (subscriber, #50428) [Link] > There is nothing that prevent distro to use multi-pass profile-guided optimization... Except for build hardware constraints. Build time is not for free. Firefox 122.0 released Posted Jan 23, 2024 22:16 UTC (Tue) by pizza (subscriber, #46) [Link] (12 responses) > There is nothing that prevent distro to use multi-pass profile-guided optimization... At minimum, it doubles the build time. But it also takes additional time/resources to run the benchmark/test/etc suites used to generate the profile data to feed back into the second pass. Plus there's the need to create/maintain the necessary environment to _run_ those suites -- presumably requiring connecting to server(s) containing representative content. This poses additional issues as some (many? most?) distros disable network access during builds. Sure, all of this is _doable_ but it's, at minimum, a considerable amount of work to set up and maintain. Plus the doubling (or so) of build times. (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) Firefox 122.0 released Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link] > (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) *chromium's 5x as long build time gives a nice friendly wave* Firefox 122.0 released Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses) In Gentoo, there's a "pho" flag to do this. Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though. Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol Firefox 122.0 released Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses) This is a strange bet coming from a m68k porter! firefox Build-Space: 38435228 Build-Time: 4843 chromium Build-Space: 13361912 Build-Time: 31142 libreoffice Build-Space: 47828428 Build-Time: 20769 llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953 Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build. Firefox 122.0 released Posted Jan 24, 2024 6:02 UTC (Wed) by roc (subscriber, #30627) [Link] Maybe not, but the important question is: do they *actually do* them? Firefox 122.0 released Posted Jan 24, 2024 9:05 UTC (Wed) by RX14 (subscriber, #123970) [Link] (1 responses) Arch Linux has been doing PGO builds of firefox ever since i first read the PKGBUILD. To hear that debian/ubuntu wouldn't, when the process is automated and they are larger distros, is strange. Firefox 122.0 released Posted Jan 24, 2024 9:44 UTC (Wed) by cyperpunks (subscriber, #39406) [Link] Fedora also do PGO build of Firefox. It's better that one build machine do double work than millions of machines doing 10% extra work.
Posted Jan 23, 2024 21:27 UTC (Tue) by pizza (subscriber, #46) [Link] (18 responses)
Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations)
This is also why Mozilla's windows binaries, when run under Wine, noticably outperform native distro-provided binaries.
(Note distros also typically perform builds with various hardening strategies enabled, which Mozilla's own builds presumably don't do. This would adversely affect performance, though many would consider this to be an acceptable trade-off)
Firefox 122.0 released Posted Jan 23, 2024 21:42 UTC (Tue) by ballombe (subscriber, #9523) [Link] (15 responses) > Mozilla does their builds in a multi-pass profile-guided optimization manner, and benchmarks have shown (historically at least) this produces faster binaries vs what the distros have done (ie using single-pass static optimizations) There is nothing that prevent distro to use multi-pass profile-guided optimization... Firefox 122.0 released Posted Jan 23, 2024 21:48 UTC (Tue) by mb (subscriber, #50428) [Link] > There is nothing that prevent distro to use multi-pass profile-guided optimization... Except for build hardware constraints. Build time is not for free. Firefox 122.0 released Posted Jan 23, 2024 22:16 UTC (Tue) by pizza (subscriber, #46) [Link] (12 responses) > There is nothing that prevent distro to use multi-pass profile-guided optimization... At minimum, it doubles the build time. But it also takes additional time/resources to run the benchmark/test/etc suites used to generate the profile data to feed back into the second pass. Plus there's the need to create/maintain the necessary environment to _run_ those suites -- presumably requiring connecting to server(s) containing representative content. This poses additional issues as some (many? most?) distros disable network access during builds. Sure, all of this is _doable_ but it's, at minimum, a considerable amount of work to set up and maintain. Plus the doubling (or so) of build times. (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) Firefox 122.0 released Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link] > (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) *chromium's 5x as long build time gives a nice friendly wave* Firefox 122.0 released Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses) In Gentoo, there's a "pho" flag to do this. Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though. Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol Firefox 122.0 released Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses) This is a strange bet coming from a m68k porter! firefox Build-Space: 38435228 Build-Time: 4843 chromium Build-Space: 13361912 Build-Time: 31142 libreoffice Build-Space: 47828428 Build-Time: 20769 llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953 Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build. Firefox 122.0 released Posted Jan 24, 2024 6:02 UTC (Wed) by roc (subscriber, #30627) [Link] Maybe not, but the important question is: do they *actually do* them? Firefox 122.0 released Posted Jan 24, 2024 9:05 UTC (Wed) by RX14 (subscriber, #123970) [Link] (1 responses) Arch Linux has been doing PGO builds of firefox ever since i first read the PKGBUILD. To hear that debian/ubuntu wouldn't, when the process is automated and they are larger distros, is strange. Firefox 122.0 released Posted Jan 24, 2024 9:44 UTC (Wed) by cyperpunks (subscriber, #39406) [Link] Fedora also do PGO build of Firefox. It's better that one build machine do double work than millions of machines doing 10% extra work.
Posted Jan 23, 2024 21:42 UTC (Tue) by ballombe (subscriber, #9523) [Link] (15 responses)
There is nothing that prevent distro to use multi-pass profile-guided optimization...
Firefox 122.0 released Posted Jan 23, 2024 21:48 UTC (Tue) by mb (subscriber, #50428) [Link] > There is nothing that prevent distro to use multi-pass profile-guided optimization... Except for build hardware constraints. Build time is not for free. Firefox 122.0 released Posted Jan 23, 2024 22:16 UTC (Tue) by pizza (subscriber, #46) [Link] (12 responses) > There is nothing that prevent distro to use multi-pass profile-guided optimization... At minimum, it doubles the build time. But it also takes additional time/resources to run the benchmark/test/etc suites used to generate the profile data to feed back into the second pass. Plus there's the need to create/maintain the necessary environment to _run_ those suites -- presumably requiring connecting to server(s) containing representative content. This poses additional issues as some (many? most?) distros disable network access during builds. Sure, all of this is _doable_ but it's, at minimum, a considerable amount of work to set up and maintain. Plus the doubling (or so) of build times. (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) Firefox 122.0 released Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link] > (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) *chromium's 5x as long build time gives a nice friendly wave* Firefox 122.0 released Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses) In Gentoo, there's a "pho" flag to do this. Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though. Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol Firefox 122.0 released Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses) This is a strange bet coming from a m68k porter! firefox Build-Space: 38435228 Build-Time: 4843 chromium Build-Space: 13361912 Build-Time: 31142 libreoffice Build-Space: 47828428 Build-Time: 20769 llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953 Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build. Firefox 122.0 released Posted Jan 24, 2024 6:02 UTC (Wed) by roc (subscriber, #30627) [Link] Maybe not, but the important question is: do they *actually do* them?
Posted Jan 23, 2024 21:48 UTC (Tue) by mb (subscriber, #50428) [Link]
Except for build hardware constraints. Build time is not for free.
Posted Jan 23, 2024 22:16 UTC (Tue) by pizza (subscriber, #46) [Link] (12 responses)
At minimum, it doubles the build time. But it also takes additional time/resources to run the benchmark/test/etc suites used to generate the profile data to feed back into the second pass. Plus there's the need to create/maintain the necessary environment to _run_ those suites -- presumably requiring connecting to server(s) containing representative content. This poses additional issues as some (many? most?) distros disable network access during builds.
Sure, all of this is _doable_ but it's, at minimum, a considerable amount of work to set up and maintain. Plus the doubling (or so) of build times.
(I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...)
Firefox 122.0 released Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link] > (I'd be willing to bet that Firefox already is already #1 or #2 of the "most resource intensive build" list...) *chromium's 5x as long build time gives a nice friendly wave* Firefox 122.0 released Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses) In Gentoo, there's a "pho" flag to do this. Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though. Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol Firefox 122.0 released Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses) This is a strange bet coming from a m68k porter! firefox Build-Space: 38435228 Build-Time: 4843 chromium Build-Space: 13361912 Build-Time: 31142 libreoffice Build-Space: 47828428 Build-Time: 20769 llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953 Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build.
Posted Jan 24, 2024 1:14 UTC (Wed) by dilinger (subscriber, #2867) [Link]
*chromium's 5x as long build time gives a nice friendly wave*
Posted Jan 24, 2024 3:27 UTC (Wed) by Trelane (guest, #56877) [Link] (7 responses)
Also, the most intensive, by far, are LibreOffice and Chromium. Firefox is no slouch though.
Firefox 122.0 released Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses) "pgo" flag. Apparently my phone wants delicious soup. And now, so do I. Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well.. Firefox 122.0 released Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses) Also, in no particular order qtwebengine, webkit-gtk, llvm and clang. Odd how half of the most painful packages to build are copies of the same codebase instead of putting that stuff into a library for reuse. Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol
Posted Jan 24, 2024 3:28 UTC (Wed) by Trelane (guest, #56877) [Link] (4 responses)
And now, so do I.
Firefox 122.0 released Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses) Oh my, is this finally the "OMG Optimize" flag for Gentoo? I just enabled it on my Firefox and the long standing tearing issue I've hard with 4K youtube videos is suddenly gone. Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well..
Posted Jan 24, 2024 18:50 UTC (Wed) by timokokk (subscriber, #52029) [Link] (3 responses)
Firefox 122.0 released Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses) Never mind, false alarm. It was just hidden lower part of the screen, didn't see it at first. Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well..
Posted Jan 24, 2024 19:07 UTC (Wed) by timokokk (subscriber, #52029) [Link] (2 responses)
Firefox 122.0 released Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses) Do you see tearing on X11 or Wayland? Which GPU/driver? Do you have hardware video decoding and also (HW_DECODED_VIDEO_ZERO_COPY) listed as enabled in about:support? Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well..
Posted Jan 28, 2024 9:49 UTC (Sun) by mokki (subscriber, #33200) [Link] (1 responses)
Firefox 122.0 released Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link] I use NVIDIA drivers 545.29.06-r1 on X11, GeForce GT 1030 display card. I don't see HW_DECODED_VIDEO_ZERO_COPY on about:support page. I also see that the codec list doesn't have any hardware decoding supported on any of the codecs. I also saw that I didn't have hwaccel use flag enabled, but enabling it didn't seem to change anything. At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well..
Posted Feb 2, 2024 18:43 UTC (Fri) by timokokk (subscriber, #52029) [Link]
At least I know what this is about... I just wouldn't have thought that I was able to run 4k 60FPS videos that smoothly without any HW support. Oh well..
Posted Jan 24, 2024 9:10 UTC (Wed) by taladar (subscriber, #68407) [Link] (1 responses)
Firefox 122.0 released Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link] I think the problem is that two of those ARE libraries and two of them are utilities. But everyone has their own set of bugfixes and tweaks, so they need to cargo their own version, rather than use the standard! I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ... Cheers, Wol
Posted Jan 24, 2024 10:27 UTC (Wed) by Wol (subscriber, #4433) [Link]
I note also that the libraries are web libraries, and the other two are under rapid development. So everyone is trying to keep up with external changes, and nobody has time to stabilise the standard ...
Cheers, Wol
Posted Jan 24, 2024 13:06 UTC (Wed) by ballombe (subscriber, #9523) [Link] (2 responses)
firefox Build-Space: 38435228 Build-Time: 4843
chromium Build-Space: 13361912 Build-Time: 31142
libreoffice Build-Space: 47828428 Build-Time: 20769
llvm-toolchain-16 Build-Space: 6428156 Build-Time: 10953
Firefox 122.0 released Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses) My eyeballs would have really appreciated some digit separation commas in those numbers, so I thought I'd offer this tweaked version for others' benefit. [1] I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own. firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s) chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s) libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s) llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s) I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes... [1] Apologies to those who think that dots should be used for this - I don't :-P [2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples. [3] Posting from mobile, so I'm at a slight disadvantage. Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build.
Posted Jan 26, 2024 12:16 UTC (Fri) by sammythesnake (guest, #17693) [Link] (1 responses)
I've assumed the units for build space and time are Kilobytes and seconds respectively, and of course any shortfall in assumptions[2] / maths / typography skills[3] are my own.
firefox Build-Space: 38,435,228 (36.7GiB) Build-Time: 4,843 (1h 20m 43s)
chromium Build-Space: 13,361,912 (12.7GiB) Build-Time: 31,142 (8h 39m 2s)
libreoffice Build-Space: 47,828,428 (45.6GiB) Build-Time: 20,769 (5h 46m 9s)
llvm-toolchain-16 Build-Space: 6,428,156 (6.1GiB) Build-Time: 10,953 (3h 2m 33s)
I'm intrigued that building Firefox takes significantly more *space* while building chromium takes far more *time*. I don't know anything of their respective build processes - presumably it makes perfect sense for those who do :-P
Are these build times for m68k? Are they *on* an m68k, or as I imagine cross-built on something more mainstream like x86-64? I wonder how much (if any) difference cross-building makes...
[1] Apologies to those who think that dots should be used for this - I don't :-P
[2] Assuming bytes came up with tiny sizes that didn't seem believable. The correctness of the assumptions isn't *super* important, given that it's just a comparison between apples and apples.
[3] Posting from mobile, so I'm at a slight disadvantage.
Firefox 122.0 released Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link] No, this is the time for the amd64 build.
Posted Jan 26, 2024 15:17 UTC (Fri) by ballombe (subscriber, #9523) [Link]
Posted Jan 24, 2024 6:02 UTC (Wed) by roc (subscriber, #30627) [Link]
Posted Jan 24, 2024 9:05 UTC (Wed) by RX14 (subscriber, #123970) [Link] (1 responses)
Firefox 122.0 released Posted Jan 24, 2024 9:44 UTC (Wed) by cyperpunks (subscriber, #39406) [Link] Fedora also do PGO build of Firefox. It's better that one build machine do double work than millions of machines doing 10% extra work.
Posted Jan 24, 2024 9:44 UTC (Wed) by cyperpunks (subscriber, #39406) [Link]
It's better that one build machine do double work than millions of machines doing 10% extra work.
Posted Jan 23, 2024 23:22 UTC (Tue) by jmalcolm (subscriber, #8876) [Link]
That is probably one of the primary motivations for releasing this .deb for Ubuntu.
As for Debian, depending on which version you choose, the official package may be quite a bit out of date. For Debian, that would be the main motivation here.
The .deb is only for Ubuntu and Debian systems. Both of the above ring true to me and fit nicely into what Mozilla is saying.
In the official release notes though, Mozilla actually recommends sticking with your distro package:
https://support.mozilla.org/en-US/kb/install-firefox-linu...
Nothing "shameful" here that I can see.
Posted Jan 23, 2024 18:17 UTC (Tue) by mb (subscriber, #50428) [Link] (4 responses)
I currently use the flatpak and it mostly works fine. It has a few OS integration issues (e.g. fonts), but it's minor enough that I can live with it.
Now, the deb will probably integrate better into the OS. At least I hope so. But what I really dislike is the proposal to prioritize the mozilla repo: >Configure APT to prioritize packages from the Mozilla repository: >Package: * >Pin: origin packages.mozilla.org >Pin-Priority: 1000
That basically means mozilla can install and override any package on my system, right?
I mean, I'm not so much worried about hostile packages. If Firefox is hostile, that's basically game over already. I'm more worried about mistakes. Or that they start updating random libraries because they need it, causing conflicts with base install packages.
What do you think? Is it worth going from flatpak to deb? Or is flatpak better?
Firefox 122.0 released Posted Jan 23, 2024 19:08 UTC (Tue) by Luban (guest, #98071) [Link] You should be able to pin only specific package(s), instead of '*', to the specific origin: https://wiki.debian.org/AptConfiguration#Prevent.2Fselect... Firefox 122.0 released Posted Jan 24, 2024 16:01 UTC (Wed) by lproven (guest, #110432) [Link] (2 responses) > But what I really dislike is the proposal to prioritize the mozilla repo This is necessary on Ubuntu, and I think that Ubuntu is the primary goal here. To explain: It's necessary because Ubuntu has a dummy Firefox .deb package which it uses to automatically prefer the Snap-packaged version. So, if you remove the snap-packaged Ubuntu version of Firefox, then add an alternative repository, then do `apt update ; apt install firefox` it will re-install the snap package. This confuses non-experts, and non-experts are the #1 target audience of Ubuntu. To bypass this, you have to set the new repo's priority as higher than the distro's own. I think it's aimed at Ubuntu because: [1] Debian includes Firefox anyway (albeit the ESR version), packaged as a .deb. [2] Linux Mint, probably the #1 Ubuntu downstream, already includes a Mozilla-codeveloped native .deb-packaged Firefox. [3] Of other Ubuntu downstreams, Linux Lite uses Chrome by default, and I *think* Zorin uses the Firefox Flatpak, as it ships several tens of gigabytes of flatpak-packaged apps as standard. (Yes, really that big: Zorin OS is the single biggest distro I've ever seen. Some other distros require as much or more disk, e.g. Deepin, because of redundant failover root partitions; Zorin OS just needs it for all the flatpak apps.) Firefox 122.0 released Posted Jan 25, 2024 4:34 UTC (Thu) by interalia (subscriber, #26615) [Link] (1 responses) > [1] Debian includes Firefox anyway (albeit the ESR version), packaged as a .deb. Eh? There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable. Firefox 122.0 released Posted Jan 25, 2024 7:23 UTC (Thu) by zaitseff (subscriber, #851) [Link] > There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable. Debian Sid has both packages (firefox and firefox-esr), but only firefox-esr can progress to Debian Testing and from there to an actual release. Bug #817954 ensures that this is the case: https://bugs.debian.org/817954
Posted Jan 23, 2024 19:08 UTC (Tue) by Luban (guest, #98071) [Link]
https://wiki.debian.org/AptConfiguration#Prevent.2Fselect...
Posted Jan 24, 2024 16:01 UTC (Wed) by lproven (guest, #110432) [Link] (2 responses)
This is necessary on Ubuntu, and I think that Ubuntu is the primary goal here.
To explain:
It's necessary because Ubuntu has a dummy Firefox .deb package which it uses to automatically prefer the Snap-packaged version. So, if you remove the snap-packaged Ubuntu version of Firefox, then add an alternative repository, then do `apt update ; apt install firefox` it will re-install the snap package. This confuses non-experts, and non-experts are the #1 target audience of Ubuntu. To bypass this, you have to set the new repo's priority as higher than the distro's own.
I think it's aimed at Ubuntu because:
[1] Debian includes Firefox anyway (albeit the ESR version), packaged as a .deb.
[2] Linux Mint, probably the #1 Ubuntu downstream, already includes a Mozilla-codeveloped native .deb-packaged Firefox.
[3] Of other Ubuntu downstreams, Linux Lite uses Chrome by default, and I *think* Zorin uses the Firefox Flatpak, as it ships several tens of gigabytes of flatpak-packaged apps as standard. (Yes, really that big: Zorin OS is the single biggest distro I've ever seen. Some other distros require as much or more disk, e.g. Deepin, because of redundant failover root partitions; Zorin OS just needs it for all the flatpak apps.)
Firefox 122.0 released Posted Jan 25, 2024 4:34 UTC (Thu) by interalia (subscriber, #26615) [Link] (1 responses) > [1] Debian includes Firefox anyway (albeit the ESR version), packaged as a .deb. Eh? There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable. Firefox 122.0 released Posted Jan 25, 2024 7:23 UTC (Thu) by zaitseff (subscriber, #851) [Link] > There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable. Debian Sid has both packages (firefox and firefox-esr), but only firefox-esr can progress to Debian Testing and from there to an actual release. Bug #817954 ensures that this is the case: https://bugs.debian.org/817954
Posted Jan 25, 2024 4:34 UTC (Thu) by interalia (subscriber, #26615) [Link] (1 responses)
Eh? There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable.
Firefox 122.0 released Posted Jan 25, 2024 7:23 UTC (Thu) by zaitseff (subscriber, #851) [Link] > There are two separate packages in Debian called 'firefox' and 'firefox-esr'. I wouldn't think they'd both be builds of the ESR version, at least in sid/unstable. Debian Sid has both packages (firefox and firefox-esr), but only firefox-esr can progress to Debian Testing and from there to an actual release. Bug #817954 ensures that this is the case: https://bugs.debian.org/817954
Posted Jan 25, 2024 7:23 UTC (Thu) by zaitseff (subscriber, #851) [Link]
Debian Sid has both packages (firefox and firefox-esr), but only firefox-esr can progress to Debian Testing and from there to an actual release. Bug #817954 ensures that this is the case: https://bugs.debian.org/817954
Posted Jan 23, 2024 19:19 UTC (Tue) by zaitseff (subscriber, #851) [Link] (1 responses)
Which is why I recompile Debian's firefox, firefox-esr and chromium packages for Ubuntu at https://ftp.zap.org.au/pub/debian/ (see the README file at that location for details), with patches and backports as needed to make this happen. I bump the epoch of the package to make it override Ubuntu's versions.
What interests me are answers to the following questions:
1. Does Mozilla provide the equivalent deb-src line for package sources? Too many upstream providers of DEB packages do NOT have such sources -- which means recreating their DEB packages is either non-trivial or impossible without access to their build infrastructure.
2. What patches from the current Debian package, if any, are included? These patches (which you can see at https://salsa.debian.org/mozilla-team/firefox/-/tree/rele...) are mostly build-time fixes, but some might be useful upstream (eg, Set-DPI-to-system-settings.patch).
Firefox 122.0 released Posted Jan 23, 2024 20:39 UTC (Tue) by ballombe (subscriber, #9523) [Link] > 1. Does Mozilla provide the equivalent deb-src line for package sources? Too many upstream providers of DEB packages do NOT have such sources -- which means recreating their DEB packages is either non-trivial or impossible without access to their build infrastructure. No they do not, as far as I can see.
Posted Jan 23, 2024 20:39 UTC (Tue) by ballombe (subscriber, #9523) [Link]
No they do not, as far as I can see.
Posted Jan 23, 2024 19:19 UTC (Tue) by ballombe (subscriber, #9523) [Link]
Posted Jan 23, 2024 21:33 UTC (Tue) by atai (subscriber, #10977) [Link] (1 responses)
I disabled and removed snap package but when upgrading software the Firefox snap package comes back. Very annoying
Firefox 122.0 released Posted Jan 24, 2024 2:19 UTC (Wed) by PengZheng (subscriber, #108006) [Link] The 5th step of the official guide should fix the priority issue: https://support.mozilla.org/en-US/kb/install-firefox-linu... echo ' Package: * Pin: origin packages.mozilla.org Pin-Priority: 1000 ' | sudo tee /etc/apt/preferences.d/mozilla
Posted Jan 24, 2024 2:19 UTC (Wed) by PengZheng (subscriber, #108006) [Link]
echo ' Package: * Pin: origin packages.mozilla.org Pin-Priority: 1000 ' | sudo tee /etc/apt/preferences.d/mozilla
Posted Jan 24, 2024 5:11 UTC (Wed) by Cyberax (✭ supporter ✭, #52523) [Link]
Posted Jan 24, 2024 7:17 UTC (Wed) by donbarry (guest, #10485) [Link] (2 responses)
If not, then their increasingly hypocritical invocations of "open source" simply expose the gulf between the claims and the practice.
Firefox 122.0 released Posted Jan 26, 2024 1:23 UTC (Fri) by higuita (guest, #32245) [Link] (1 responses) The digital sign is there to protect everyone, that avoids normal people from installing trojan extensions. Now we have few attacks using it exactly because of the required signature, but in the past, there were many fake extensions for firefox... with most extensions being opensource, someone can pull the source and adding some extra code and name it the same or similar and will always catch someone. There are ways to install unsigned packages in official firefox, but those require several extra steps, so not something that people would do by accident. Or use the developer versions or nightly and change the xpinstall.signatures.required setting, those can allow unsigned extensions, usually for developer testing, but if you want to take the risk, use those versions Firefox 122.0 released Posted Jan 29, 2024 15:10 UTC (Mon) by intelfx (subscriber, #130118) [Link] > There are ways to install unsigned packages in official firefox, but those require several extra steps, so not something that people would do by accident Are there? I'm genuinely interested.
Posted Jan 26, 2024 1:23 UTC (Fri) by higuita (guest, #32245) [Link] (1 responses)
There are ways to install unsigned packages in official firefox, but those require several extra steps, so not something that people would do by accident. Or use the developer versions or nightly and change the xpinstall.signatures.required setting, those can allow unsigned extensions, usually for developer testing, but if you want to take the risk, use those versions
Firefox 122.0 released Posted Jan 29, 2024 15:10 UTC (Mon) by intelfx (subscriber, #130118) [Link] > There are ways to install unsigned packages in official firefox, but those require several extra steps, so not something that people would do by accident Are there? I'm genuinely interested.
Posted Jan 29, 2024 15:10 UTC (Mon) by intelfx (subscriber, #130118) [Link]
Are there? I'm genuinely interested.
Posted Jan 24, 2024 17:26 UTC (Wed) by traxus (guest, #119726) [Link]
Copyright © 2024, Eklektix, Inc. Comments and public postings are copyrighted by their creators. Linux is a registered trademark of Linus Torvalds
LWN